Reviving lost shadows: investigating the habitat ecology of the rediscovered hispid hare (Caprolagus hispidus) in Nepal

Study area

The research was conducted in Chitwan National Park (CNP), the first national park in Nepal and a designated UNESCO World Heritage site (see Fig. 1). CNP, along with its buffer zone, is situated in the southern part of Central Nepal, spanning across Chitwan, Nawalparasi, Parsa, and Makwanpur districts. The park’s coordinates range from N27°20′19″ to 27°43′16″ longitude and E83°44′50″ to 84°45′03″ latitude, covering an expanse of 952.6 km² in the Rapti Valley within the Siwalik physiographic region. The buffer zone, extending from N27°28′23″to 27°70′38″ longitude and E83°83′98″ to 84°77′38″ latitude, encompasses an area of 729.37 km². CNP’s elevation varies from 100 m in river valleys to 815 m in the Siwalik hills (Bhuju et al., 2007). The park is rich in biodiversity, hosting an impressive array of wildlife, including almost 68 species of mammals, over 576 species of birds, 49 species of reptiles and amphibians, 120 species of fishes, and numerous invertebrates, all of which play a crucial role in the park’s ecological processes (CNP, 2015). Among the charismatic species, Chitwan National Park is home to 128 Royal Bengal Tigers, the highest number among all other national parks (Pokharel, 2022). Moreover, the park is home to the second-largest population of the Greater One-horned Rhinoceros in the subcontinent, ranking just below Kaziranga National Park in India (Mandal, 2022). The dominant grass species in the grasslands include Saccharum munja, Saccharum bengalense, and Imperata cylindrica (Khadka et al., 2017).

Data collection

Before an extensive field survey, a preliminary key informant survey was conducted in January 2023. The participants included representatives from the national park (specifically rangers, n = 3), members of the buffer zone management committee (n = 2), and nature guides (n = 3). The key informants were chosen based on their prior experience working in the national park. To assist them in the process, detailed historical distribution maps (color-printed) were provided during interviews. Following this, the research team conducted an initial field trip to locations recommended by the key informants, aiming to understand the terrain and plan for the comprehensive field survey. Researchers recorded GPS locations using handheld Garmin devices whenever indirect evidence of the hispid hare, such as pellets, dens, or grass cutting, was encountered.

The extensive field visit was conducted during the cold winter, specifically from 28 January to 13 February 2023. This temporal preference was guided by the sparser vegetation cover in the tropic region in this season (Safford, 2004; Sanusi et al., 2013; Neupane et al., 2022). The spare vegetation cover enhances the chances of spotting hispid hare pellets, thus mitigating potential biases linked to undetectability issues (Sadadev et al., 2021; Dhami et al., 2023a; Dhami et al., 2023b). Non-intrusive molecular methods serve as more compelling evidence for confirming species’ presence (Buglione et al., 2020b; Buglione et al., 2020a). However, since the hispid hare is nocturnal and elusive, direct sightings are rare, making pellet analysis a practical alternative for confirming its presence. The size and shape of pellets are recognized as reliable indicators for species identification (Aryal et al., 2012; Chand, Khanal & Chalise, 2017). Hispid hare pellets can be distinguished from those of the rufous-tailed Indian hare (Lepus nigricollis ruficaudautus) by their unique larger size and flattened, rounded shape (Dhami et al., 2023a). In contrast, Indian hare pellets are generally smaller, darker, and exhibit an elliptical form with a pointed end (Refer to Sadadev et al., 2021 for photographic evidence).

We used the transect method for our survey, laying down a long linear transect of 500 m. Adjacent transects were spaced 200 m apart to minimize extraneous variation within the location and minimize the possible overlapping of pellets (Dhami et al., 2023a). We placed alternate strip transects along this main transect, each 20 m long and 2 m wide (Sadadev et al., 2021). These strip transects were positioned at 100-meter intervals along the main transect, as shown in Fig. 2. A team led by the first author and experienced field guides with a background in wildlife research and monitoring systematically searched along the strip transect for indirect signs of the hare’s presence (pellets, dens, and grass cutting). This method was used because direct sightings were difficult due to the hare’s nocturnal and elusive nature (Bell, 1987; Bell, Oliver & Ghose, 1990). Whenever the indirect signs of a hispid hare were encountered, a circular plot of radius 1.78 m was established with detected signs at the center (Gyawali, 2003). Subsequently, another circle with the same radius was established in a random direction, 100 m away from the center of the initial circle with detected signs (Khulal et al., 2021; Neupane, Chhetri & Dhami, 2021; Neupane et al., 2022). These circular plots served as standard habitat samples, independent of whether the hispid hare was present. If any signs of a hispid hare were detected within the circles, a value of 1 (“utilized plots”) was assigned; otherwise, if no evidence was found, a value of 0 (“habitat availability plot”) was assigned.

Based on the literature review (Aryal et al., 2012; Chand, Khanal & Chalise, 2017; Dhami et al., 2023a; Khadka et al., 2017; Nath & Machary, 2015; Oliver, 1984; Rastogi, Raj & Chauhan, 2020; Sadadev et al., 2021; Tandan et al., 2013) and initial on-site survey we recorded nine predictor variables that could potentially influence the habitat preferences of hispid hare across all circles of both types (“used plots” and “availability plots”) (Table 1). Firstly, the water shapefile was extracted from DEM (Digital elevation model) with a resolution of 12.5 m and a Landsat Image 8 (USGS, 2021), and the Road/Path/Fireline shapefile was extracted from the Open street map (OSM, 2021). Then, the distance to the nearest water source and Road/Path/Fireline was calculated using a “Nearer function” in ArcGIS 10.8. version for the presence locations (geographic coordinates) of the hispid hare. Habitat-type categorization and recording of the presence/absence of anthropogenic disturbance, invasive species, and Predator was based on personal observations (Koirala et al., 2020; Khulal et al., 2021). For determining the presence of predators such as Felis chaus and Canis aureus, we relied on identifying their fresh signs, specifically pugmarks and scats that were estimated to be 1–3 days old (Khatoon et al., 2019). Similarly, we visually approximated ground cover using a circular bamboo frame (radius: 1.78 m) intersected by two diagonal sticks (Roshetko et al., 2002). We estimated the total percentage of grass species coverage by adding the proportion of coverage in each quadrant within the circular frame. The dominant plant species were identified by determining the species with the highest percentage coverage recorded within the circle (Dhami et al., 2023a). The ground condition was evaluated by considering soil moisture’s presence or absence (Table 1).

Data analysis

The distribution pattern of the hispid hare was first determined using the variance-to-mean ratio (S²/a). The calculation involves S² = $\frac{1}{n}\sum {\left(x-a\right)}^2$, where x denotes the number of pellet groups (signs) per sampling unit, a is the mean of the x values, and n represents the number of sampling units (Odum, 1971).

If (S²/a) = 1, i.e., a random distribution,

If (S²/a) < 1, i.e., a uniform distribution,

If (S²/a) > 1, i.e., a clump distribution.

Then, a test for multicollinearity was conducted on the selected independent habitat predictors using the variance inflation factor (VIF) with the “Faraway” package (Boomsma, 2014) in R x 64 4.0.3 (R Core Team, 2020). All selected habitat predictors were included in the final analysis, as none exhibited multicollinearity, with tolerance values exceeding 0.1 and VIF values below 10 (Bowerman & O’connell, 1990). Generalized linear models (GLMs) utilizing a binary distribution were employed to assess the probability of detecting hispid hare with distinct habitat predictors outlined in Table 1. Among the nine habitat predictors, the nearest distance to a water source (WD) and Road/path/firelines were treated as a continuous variable. In contrast, the remaining predictors were considered categorical variables in the analysis. We employed the ‘dredge’ function within the “MuMIn” package (Barton, 2009) to execute the global model, generating all potential models. These models were subsequently ranked using Akaike’s information criterion with correction (AICc) for small samples, with the most effective or dominant models identified by their lower AICc values (Barton & Barton, 2020). The final model-averaged coefficients were obtained by averaging the top candidate models (with delta AIC ≤ 2) (Burnham & Anderson, 2001). This averaging process was performed using the “AICmodavg” package (Mazerolle & Mazerolle, 2017). The predictive efficacy of the best-fitting model was assessed using the area under the curve (AUC) of the receiver operating characteristics (ROC) values ranging from 1 to 0, using the R package “ROCR” (Sing et al., 2005) (Fig. 3). Discrimination performance was considered acceptable for values between 0.7 and 0.8, excellent for values between 0.8 and 0.9, and superior for values exceeding 0.9 (Hosmer, Lemeshow & Lemeshow, 2000).

Distribution pattern of hispid hare

Hispid hare presence was predominantly recorded in Sukhibhar grassland (68.5%), followed by Reu Khola Phanta I (17.5%) and Reu Khola Phanta II (14%) (Fig. 1). The overall distribution exhibited a clumped pattern with a variance to mean ratio >1, i.e., 2.36.

Influencing variables and probability of hispid hare occurrence

Among the binary logistic regression models predicting the likelihood of detecting the hispid hare, the model that includes the combined effects of habitat type and distance to water emerged as the most suitable. This model achieved the lowest Akaike Information Criterion (AIC) value of 160.43 and the highest model weight of 0.36, indicating superior performance compared to all other models (Table 2). Further, the best-fit model demonstrated exceptional performance in explaining hispid hare detection probability, with a 0.803 (80.3%) receiver operating characteristic (ROC) curve area and an impressive 0.805 (80.5%) accuracy (Fig. 3).

Among the nine habitat predictors studied, tall grassland habitat type [HT]1 (β = 1.6009531, S.E = 0.7732176, P = 0.038405), short grassland habitat type [HT]2 (β = 2.6622712, S.E = 0.7349409, P = 0.000292) and water distance [WD] (β = −0.0402213, S.E = 0.0094737, P = 2.18e−05) showed significant links to hispid hare detection (Table 3). The model suggests that an increase in tall and short grassland positively correlates with the likelihood of hispid hare detection. Conversely, as water distance increases, the detectability of hispid hare decreases, and vice versa (Table 3 and Fig. 4).

Spatial distribution pattern of hispid hare

Until the 1980s, the hispid hare was historically documented in three protected areas in Nepal, namely CNP, ShNP, and BNP (Oliver, 1980). By the 1980s, recorded sightings of the species were limited to BNP and ShNP, raising concerns about its distribution and population status (Tandan et al., 2013). Surprisingly, an individual hispid hare was rediscovered in Sukhibhar grassland within CNP on 30 January 2016, during a targeted survey focused on grassland birds, particularly the Bengal florican (Houbaropsis bengalensis) (Khadka et al., 2017). The rediscovery of the hispid hare in Sukhibhar grassland within CNP offers a beacon of hope for conservationists. Once thought to have expired from CNP, this revelation suggests a broader habitat range than initially recognized, enhancing our understanding of the species. This underscores the significance of thorough and targeted surveys in conservation efforts, encouraging collaborative efforts across protected areas to ensure the long-term survival of this elusive species. In our study, Sukhibhar grassland stands out as a central hub for distributing the hispid hare, followed by Reu Khola Phanta I and Reu Khola Phanta II. Our results are consistent with the findings of Aryal et al. (2012), Chand, Khanal & Chalise (2017), and Dhami et al. (2023a), providing further evidence that the hispid hare species favors the grassland system as an integral part of its habitat. The choice of grasslands is driven by factors such as the availability of food, shelter, and a preferred microhabitat for survival and reproduction (Dhami et al., 2023a). The hispid hare’s strong affinity for grassland ecosystems is a key factor contributing to its clumped distribution. Sadadev et al. (2021) found a clumped distribution of hispid hare in both burned and unburned plots within ShNP, mirroring the results of our study. These results shine a light on the important factor that species with specific habitat requirements often exhibit non-random distribution patterns (Underwood, 1978). Furthermore, CNP harbors the highest number of one-horned rhinoceros (Rhinoceros unicornis) (n = 694) in Nepal, a grassland-dependent species with similar feeding preferences as hispid hare (Subedi et al., 2017). Thus, a study focusing on species interaction is recommended to better understand the dynamics and dependencies within the ecosystem.

Influencing variables and probability of hispid hare occurrence

We observed a positive correlation between the likelihood of detecting hispid hare and the presence of both tall and short grasslands. Aryal et al. (2012) also found that the hispid hare predominantly used tall grasslands (2–6 m) before and after a fire event. Additionally, pellets were discovered in open areas with short grass cover (<2 m), with a notable increase after the fire. This suggests a flexible habitat hispid hare use, highlighting their adaptability to varying grassland conditions. The observed patterns in habitat preference may be influenced by factors such as food availability, shelter, and microhabitat requirements (Bell, 1987; Sadadev et al., 2021; Dhami et al., 2023a). Numerous global studies on hare species, such as the European brown hare (Lepus europaeus), have similarly indicated that grasslands and fallow lands serve as primary habitats for hare (Smith et al., 2004; Sliwinski et al., 2019). This strong habitat preference may stem from the fact that tall grasslands and vegetation act as resting places (Smith et al., 2004; Sliwinski et al., 2019), offering essential shelter and cover against predators (Tapper & Barnes, 1986; Jennings et al., 2006). This protective environment can enhance reproductive success by reducing the risk of predation on offspring (Aryal et al., 2012; Sadadev et al., 2021). Furthermore, the presence of the hispid hare was notably high in habitats dominated by species such as Saccharum spontaneum and Imperata cylindrica (Yadav, 2006; Dhami et al., 2023a), both of which were identified as common components of the hispid hare’s diet in a study by Aryal et al. (2012) conducted in ShNP. During the monsoon season in Assam, India, hispid hare are noted to inhabit forested areas due to waterlogging of the grasslands (Ghose, 1978). However, there is no reported evidence of similar behavior in Nepal (Aryal et al., 2012; Tandan et al., 2013; Chand, Khanal & Chalise, 2017; Sadadev et al., 2021; Dhami et al., 2023a).

After the establishment and growth of Nepal’s protected areas network, many have served as tourist attractions for many years (Heinen & Thapa, 1988), contributing millions of dollars annually to both the national and local economies (Baral et al., 2020). The rise in tourism has resulted in the disruption of wildlife activities, increased stress levels, and the loss or modification of habitats. For instance, (Cosgriff, 1997) observed alterations in the behavior of greater one-horned rhinoceros in response to tourism activities, with animals avoiding typical habitats and seeking refuge in isolated locations within CNP. Similarly, Dhami et al. (2023a) demonstrated a negative correlation between the probability of hispid hare detection and anthropogenic disturbance in ShNP, indicating the direct impact of human activities on the well-being of the hispid hare population and other wildlife.

A negative correlation is evident between the distance from water sources and the detection of hispid hare, indicating that the likelihood of detecting hispid hare decreases as the distance from the water source increases. This aligns with what was found in a prior investigation by Scharine et al. (2011), which documented comparable trends in other lagomorph species. Tandan et al. (2013) also, reported that the hispid hare exhibits a broad distribution, particularly in areas closer to water sources. This could be attributed to the winter season, where water scarcity might compel the hispid hare to remain in areas close to water sources for hydration. Additionally, water bodies often sustain abundant vegetation, attracting various plant species and thereby offering ample foraging opportunities during the dry season (Neupane, Chhetri & Dhami, 2021; Regmi et al., 2022). In contrast to our findings, Nath & Machary (2015) reported a high concentration of hispid hare pellet piles in areas characterized by dry ground conditions and at a distance from water sources. Consistent with this, the presence of hispid hare signs was more commonly found in areas distant from water sources after a fire, especially when there was new growth grass, compared to situations with only unburned grass (Aryal et al., 2012). This suggests that the availability of new grass, characterized by its increased water content, allows hispid hare to extend their habitat beyond water sources following fires (Aryal et al., 2012; Yadav et al., 2008).